Typically one would perform an Analysis of Variance ( anova1) on the actual dataset, not on your summary statistics. This will test the null hypothesis of equal mean returns vs. the alternative that at least one day's mean return is different. If you want a pairwise comparisson you can then pass the stats output to the multcompare function.
[p,tbl,stats] = anova1(data);
c = multcompar(stats);
If your data (returns) are non-normal then use the kruskal-wallis test, similar concept but uses the medians.
